基于神经网络的颗粒图像边缘混合检测方法

Hybrid Edge Detection of Globular Material Using Neural Network

针对颗粒图像的特点，提出一种基于神经网络的边缘混合检测方法。该方法包含边界候选象素提取和神经网络边缘检测两部分，神经网络由用于图像信息压缩与图像信息编码的自组织竞争子神经网络（ＡＳＣＳＮＮ）和用于获取图像边缘矢量信息的基于径向函数子神经网络（ＲＢＦＳＮＮ）组成。实验结果表明，该方法分割颗粒图像得到的边缘图像封闭性好、边界描述真实，适用于堆积颗粒物料图像的边缘检测。

This paper proposes an edge detection of globular material using neural networks. It includes the extraction of edge pixel candidates and edge detection using neural networks. The neural network consists of self-organizing competitive subnet used for image compressing and image encoding, and the radial basis function subnet used for deducing edge vectors. The tests showed that images segmented by this method have good edge closedness and true edge and are suitable for the segmentation of the cumulate particle image.